Voting Rights and Reproductive Rights for Artificial Intelligences

Published on August 5, 2012 by

Let’s posit that at some point, we develop self-aware artificial intelligences that are at least as capable as humans (by whatever criteria is generally agreed upon.)  We could use the character of Data from Star Trek: The Next Generation as an example of the level of intelligence and self-awareness. I think the general consensus among science fiction fans is that such an AI should probably be considered a “person” and should have rights, rather than just being property.

I tend to agree with that consensus, but I’ve run into some potential problems with granting human-equivalent rights to AIs.

Imagine that AIs are considered persons under U.S. law, with rights equal to those of a natural-born citizen (if the AI was originally turned on in the U.S.).  So, after being “alive” for eighteen years, an AI could register to vote in elections.  And at age thirty-five, an AI would be eligible to become President.

At least since the Civil Rights era, American courts have not looked kindly on laws restricting the right of certain categories of people to reproduce. A law restricting the number of children someone could have based on their race would (rightly) be considered anathema.  So laws limiting the reproductive rights of AIs would be unconstitutional.

What if an ambitious seventeen-year-old AI decided it really, really wanted to be president, so it created 250,000,000 copies of itself — each of which would be considered a person. Eighteen years later, the original AI is old enough to become President, and the 250,000,000 copies (which will have diverged somewhat based on their experiences, but still have similar personalities) will be old enough to register to vote.  The AI wins the election even if all the eligible human beings vote against it.

That’s an extreme case, obviously.  Other AIs might spawn in huge numbers to vote against that ambitious one.

But the point is this: combining the right to vote with the right to essentially unlimited reproduction for AIs means making humans an irrelevant minority in the United States.

I’m not sure that “I, for one, welcome our new AI President” is the right response.

The main problem is that our Constitution and laws were not set up to deal with artificial intelligences.  To some degree they were set up granting rights to white landowning males, and since then we have several precedents of extending those rights to other persons.  But when dealing with a truly unprecedented kind of “person,” should such precedents necessarily be followed?

Filed under: General

5 comments on “Voting Rights and Reproductive Rights for Artificial Intelligences”

  1. John Murphy says:

    That is an interesting thought. What does reproduction actually mean in this context, I wonder: human reproduction is at best an extremely imperfect copy, which is implicit in the concept. Also, modern computer programs are multi-threaded affairs: parallelism is inherent in so much of what we do under the hood that it is largely transparent. An AI would almost certainly have multiple, probably uncountably many processes comprising its “personhood”, being spawned and culled as necessary to perform tasks. One could make an argument on those two bases, then, that all 250 million processes are in fact the same person, now suffering from a self-inflicted personality disorder.

    • Carl Duzett says:

      That makes the most sense to me. But if their life experiences have the potential to diverge, somebody somewhere will be quite upset at them all being considered the same person.

  2. Easy solution:

    Anti-cloning laws.

  3. Mel Windham says:

    A very funny yet thought provoking post.

    We’re doomed, aren’t we? Well, I for one would prefer an AI Overlord rather than Kodos/Kang from the Simpsons.

    Related: you may like the recent xkcd “What If” article that describes what a robot apocalypse would look like if it happened today…

    http://what-if.xkcd.com/5/